World records in athletics capture the attention of both the general public and the academic community. Athletics records provide a wealth of quantifiable data, which represent the pinnacle of human ability at the time of the record. This research discusses modelling the progression of athletics world records using a compound distribution model which is found in actuarial statistics. Compound distributions allow for the simultaneous modelling of the frequency of record breaking and of the amount by which the record is improved at each record-breaking performance. A number of assumptions accompany the use of the compound distribution model. This research evaluates the athletics data for adherence to these assumptions, before carrying out distribution-fitting and goodnessof-fit testing. Forecasts of future performance are obtained by means of a simulation method. The forecasts obtained from the compound model are compared to forecasts obtained from non-linear models which have been proposed in literature. This research focuses on short-term forecasts, however, a modification to the compound model which allows for longer-term forecasts is also illustrated.